Now even nixos-18.09-small has been stuck for a week. If there isn’t any low-hanging fruit left I think the way NixOS is developed must drastically change to be viable.
I think we can definitely do a lot to make some of these tests less flaky. Especially, the installer tests have had issues that are usually random. But, we actually do want things to get stuck when things are legitimately broken.
This particular hang wasn’t about tests themselves but about the build farm “maintenance”. TL;DR: you can’t well automatize everything – some humans still have to watch things and fix them when needed.
For those of you who were brought here by Google looking for help updating NixOS:
(I know this thread is only tangentially related, but this is where google keeps corralling that search query)
Here is my update/upgrade process…
Initially, using whatever channel you want:
sudo nix-channel --add https://nixos.org/channels/nixos-unstable nixos
Then to update:
sudo nix-channel --update nix-channel --update sudo nix-env --upgrade nix-env --upgrade sudo nixos-rebuild switch --upgrade
This was suggested to me on the IRC when I started asking questions about “why doesn’t setting xyz in my configuration.nix cause my system to upgrade?”
I was very surprised to learn that most of those settings didn’t do what I had imagined based on their names. This process might not be the “proper way” but it works well for me.
There are ux problems related to channels. The replacement is work in progress: https://gist.github.com/edolstra/40da6e3a4d4ee8fd019395365e0772e7
To revive this thread: Why hasn’t trunk-combined be evaluated over the last 5 days? Here is the last eval: https://hydra.nixos.org/build/86749887
There seems to have been an introduced an error in the kerberos test. @pbogdan have supplied a fix which should hopefully do the trick:
Thanks! Is there some way to find out about failed hydra evals?
As in be notified when it happens? Not that I know of. Perhaps if you have an hydra account, but that seems pretty restrictive at the moment.
Is there a manual way to check?
Last checked timestamp in the
Evaluation tab, so when that doesn’t result in a new eval there’s probably something wrong. The errors for the last check can be found in the
Evaluation errors tab.
For some reason the last check didn’t go through either, but the error blocking for the last days is gone at least.
Thanks, that’s good to know for the next time.